An investigation on the applications of advanced Infrared Spectroscopy, Spectral Imaging and Machine Learning for Polymer Characterization, including microplastics

The degradation of polymers and the quantitative and qualitative analysis of microplastics have become significant issues in the fields of materials science and environmental health. This study integrates advanced spectral imaging techniques, chemometrics, and machine learning to identify and characterize polymer degradation processes while also enhancing the accuracy of microplastic detection and improving standardization of the process. The thesis first explores how portable Attenuated Total Reflectance Fourier Transform Infrared Spectroscopy (ATR-FTIR), optical photothermal infrared (O-PTIR) spectroscopy, and Raman imaging can be combined with multivariate analysis (principal component analysis) and classification models (partial least squares discriminant analysis, support vector machines, and random forest classification) to identify and characterize the different degradation stages of polymers during UV-induced aging. Notably, O-PTIR spectroscopy (offering submicron spatial resolution) revealed the heterogeneity of polyurethane and polystyrene surfaces and can be used to study degradation depth, enabling the development of a highly accurate model capable of distinguishing different UV-induced degradation stages. Building on high-resolution O-PTIR, spectral information of materials on a smaller scale (around 500 nm) can be obtained, a novel microplastic (MP) detection framework was developed, integrating optimized sample pretreatment (such as high-temperature filtration and alcohol treatment to reduce non-polymer debris) with O-PTIR imaging technology for detection at specific infrared wavenumbers. An SVM classifier trained on four key wavenumbers demonstrated high accuracy in distinguishing nylon microplastic particles from other materials, significantly reducing analysis time compared to full-spectrum imaging and enabling characterization of each particle in the image. Using this framework, the study quantified microplastic release from nylon tea bags, with each tea bag releasing approximately 106 MP particles under typical usage conditions. Additionally, by identifying the limitations of current microplastic identification methods, a more reliable matching method combining O-PTIR spectra with a commercial Polymer Library and human visual confirmation was developed. This led to the finding that common laboratory solvents used in sample preparation may introduce diverse microplastic contaminants (primarily polyamide fragments). Extending the method to consumer products, this work demonstrated that salt grinder heads release far more MPs than the intrinsic MP content of the salt itself, highlighting a previously overlooked pathway of direct dietary exposure. These findings indicate that combining advanced spectral imaging technology with chemometric analysis can effectively advance polymer degradation research. The research outcomes provide new insights into understanding polymer aging mechanisms and establish faster, standardized microplastic detection methods, which are crucial for enhancing polymer durability, guiding recycling strategies, and assessing environmental and health risks posed by plastic fragments.